A comparison with lexicosemantic rules

The main insight for the comparison betweenCGand lexicosemantic rules (LSR) is that changes in argument structure are crucially semantic. In aLSRapproach, the argument structure is taken to be uniquely predictable from the semantic representation of the verb. Therefore, if a verb is correctly used with different argument structures, this reflects differences in the semantic representation of

the verb. On the contrary, in the approach proposed by Goldberg, argument structure is (relatively) independent from the verb. These two approaches are compared with respect to a series of points with the aim to compel the reader that theCGapproach is more plausible than the LSRapproach. These points are taken in turn in what follows.

Implausible verb senses are avoided

Consider again example (6.1) and compare it with ‘John sneezes’. The latter is an intransitive construction —in fact, ‘sneeze’ is an intransitive verb. The former is a ditransitive construction. Since we have two different argument structures, a

LSRapproach predicts different semantic representations of ‘sneeze’, one for the transitive, ‘sneeze1’, and another for the ditransitive, ‘sneeze2’. This distinction would make sense if there were a language with two different words, one with the meaning of ‘sneeze1’, and another with the meaning of ‘sneeze2’. However, this seems very implausible.

Under aCGapproach, differences in meaning are due to different construc- tions. So, in both sentences, ‘sneeze’ has the same meaning, and the differences in meaning of the whole sentence can rather be attributed to the different con- structions.

Circularity is avoided

Goldberg points out a circularity embedded in the LSR approach. To begin with, she presents eight sentences with the verb ‘kick’, all of them with different argument structure.3 _{Some of them are:}

(6.2) Patrick kicked the ball,

(6.3) Patrick kicked his foot against the chair, (6.4) Patrick kicked Mary a ball.

She claims that both the assertion that ‘kick’ has a particularn-argument mean- ing and the explanation that ‘kick’ has n arguments come from the fact that ‘kick’ occurs in that particularn-argument structure. To make it clearer, the circularity is that ‘kick’ is claimed to have a n-argument meaning because it occurs withnarguments, and at the same time it is claimed that ‘kick’ occurs withnarguments because it has an-argument meaning.

Since in the CGapproach the arguments are associated directly to the con- struction, the circularity is avoided. For CG does not claim that ‘kick’ has a

n-argument meaning because it occurs withnarguments.

Explaining non-compositional cases

The principle of compositionality is criticized by Goldberg by means of two phe- nomena: the Dutch impersonal passive construction and the way construction in English. I will only discuss the Dutch impersonal passive construction.

The Dutch impersonal passive construction: Consider the following examples: (6.5) (a) Er werd gelopen,

There was run,

(b) *?Er werd naar huis gelopen, There was run home,

(c) Er werd voordurend naar huis gelopen, There was constantly run home.

The constraint seems to be that the impersonal passive is allowed for atelic situations (like those in (6.5a, c)), and not for telic situations (like the one in (6.5b)). This implies that this is a constraint on the whole expression, and since the verb (‘gelopen’) is the same in all cases, it follows that the constraint cannot be expressed as a lexical constraint. Postulating a telic and an atelic meaning of the verb ‘gelopen’ would allow us to keep this constraint as a lexical constraint. However, such a move would not allow us to explain the behavior of the auxiliaries ‘zijn’ and ‘hebben’ (the former used with telic and the latter with atelic verbs). They depend exclusively on the aspect of the verb, independent of the aspect of the whole sentence. Thus, a telic sense of (6.5a) is disallowed, which makes less plausible the possibility of postulating different meanings of the verb ‘gelopen’. Goldberg claims correctly that the principle of compositionality implies that the meaning of an expression is determined only by the meaning of the lexical units that occur in it. There is no other kind of meaning that contributes to the meaning of the expression. (If we said that the meaning of an expression is determined also by the meaning of a subexpression we would not add anything, since the meaning of that subexpression is in turn determined by the meaning of the lexical elements that occur in it.) The composition rules are, in this sense, empty with respect to its contribution to the meaning of the expression —they just sort ofcombinemeanings, but they do notaddmeaning.

In trying to make sense as to how the Dutch impersonal passive construction bears upon compositionality, we can think that the aspect of a situation is an integral part of the meaning of a sentence. As an element of meaning, compositionality implies that aspect should be determined at the lexical level —and there is no other option that ascribing it but to the verb.4 _{Here is where} the problem lies, since aspect seems to be required distinctly for the verb and for the whole sentence. Examples (6.5a-c) motivate a change in the aspect of the whole sentence, whereas the auxiliary selection phenomenon motivates a fixed verb aspect. Thus, where does the aspect of the whole sentence come from?

Under theCGapproach the answer is: the aspect of the whole sentence comes from the aspect of the construction. Goldberg has the following alternative claim 4_{This chain of reasoning works out as long as we consider the aspect of a sentence not to}

be a composition of other sort of meanings. It is true, though, that it sounds strange to say that aspect is made up from other things that are not aspectual meanings.

to the principle of compositionality —which she claims to be a weak form of compositionality: ‘The meaning of an expression is the result of integrating the meanings of the lexical items into the meanings of constructions’.5

Supportive evidence from sentence processing

Goldberg upholds the hypothesis that the use of the same meaning of a verb in different argument structures do not show the same processing effects that cases of real lexical ambiguity do. Therefore, it seems plausible that we can measure the LSR thesis that a difference in argument structure implies a difference in verb’s meaning. For, in the face of a couple of sentences with different argument structure like:

(6.6) Bill loaded the truck onto the ship, (6.7) Bill loaded the truck with bricks,

the phenomenon predicted byLSR is that of lexical ambiguity, and not that of same core meaning occurring in different argument structures. However, Carlson and Tanenhaus found that the processing times of sentences like (6.6-6.7) are different from sentences with real semantic ambiguity, like:

(6.8) Bill set the alarm clock onto the shelf, (6.9) Bill set the alarm clock for six.

Their hypothesis is this: If a reader or hearer initially selects an inappropriate sense of an ambiguous word like ‘set’, the processing load will increase. On the other hand, if an inappropriate constructional use is selected, the reanalysis will be relatively cost free since the sense of the verb remains constant and the verb’s participant roles are already activated. The empirical evidence provided by Carlson and Tanenhaus, summarized in [5, p.18] seems to support this thesis.